DIRECT MEDICAL IMAGE TO SIMULATION USING AUTO-SEGMENTATION AND POINT CLOUD-BASED CFD

Area of Science:

• Computational medicine
• Biomedical engineering
• Medical imaging

Background:

• Cardiovascular disease (CVD) is a major global health concern.
• Current image-based computational methods for CVD analysis face challenges in automation and require specialized expertise.
• Digital twins offer potential for mitigating CVD impact but require efficient simulation pipelines.

Purpose of the Study:

• To present a novel framework for accelerating computational fluid dynamics (CFD) simulations for cardiovascular digital twins.
• To overcome limitations of manual segmentation and domain expertise in current image-to-simulation workflows.
• To enable rapid, direct CFD simulations from patient medical images.

Main Methods:

• Developed a framework using immersogeometric analysis for direct point cloud-to-CFD simulations.
• Integrated advanced auto-segmentation techniques to extract patient-specific geometries as point clouds.
• Validated the framework with benchmark flow problems and applied it to patient-specific cardiovascular images.

Main Results:

• Demonstrated the ability to perform rapid CFD simulations directly on point clouds derived from medical scans.
• Successfully applied the framework to patient-specific cardiovascular imaging data.
• Showcased accelerated image-to-simulation pipeline for cardiovascular digital twins.

Conclusions:

• The proposed framework significantly accelerates the process of creating cardiovascular digital twins.
• This approach enhances the tractability and accessibility of advanced CVD computational analysis.
• The method holds promise for improving patient-specific cardiovascular disease assessment and treatment planning.